Neuroinformatics - Universitetet i oslo · Neuroinformatics _____ Volume 1, 2003 Introduction The...

0 0 0 1 0 1 0 1 1 1 0 0 1 0 11 1 0 0 1 0 1 0 0 1 1 1 0 1 00 1 0 1 0 1 0 0 0 1 0 1 0 0 10 0 1 1 0 0 1 1 1 0 1 0 1 0 1 01 0 1 0 1 0 1 0 0 0 1 1 0 0 1 10 1 0 1 0 0 0 1 0 1 1 0 0 1 0 00 1 0 1 0 0 0 1 0 1 0 1 1 0 1 0 01 1 1 0 0 1 0 0 0 1 0 1 1 0 0 0 01 0 1 0 1 0 0 1 0 1 0 1 0 1 0 1 00 0 0 1 0 1 0 1 1 1 0 0 1 0 11 1 0 0 1 0 1 0 0 1 1 1 0 1 00 1 0 1 0 1 0 0 0 1 0 1 0 0 11 0 1 0 1 0 1 0 0 0 1 1 0 0 1 10 1 0 1 0 0 0 1 0 1 1 0 0 1 0 0

NeuroinformaticsNeuroinformaticsISSN 1539–2791 Volume 1 • Number 2 • 2003

Editors

Giorgio A. Ascoli

Erik De Schutter

David N. Kennedy

IN THIS ISSUE

Neuroscience Data andTool Sharing: A Legal andPolicy Framework forNeuroinformatics

NeuroSys: A SemistructuredLaboratory Database

Neuroanatomical TermGeneration andComparison BetweenTwo Terminologies

Sorting Functional Classesof Evoked Potentials byWavelets

Public Policy Forum

Neuroinformatics JournalCopyright ©Humana Press Inc.All rights of any nature whatsoever are reserved.ISSN 1539-2791/03/01:149–166/$20.00

AbstractThe requirements for neuroinformatics to make

a significant impact on neuroscience are not sim-ply technical—the hardware, software, and pro-tocols for collaborative research—they also includethe legal and policy frameworks within which pro-jects operate. This is not least because the creationof large collaborative scientific databases ampli-fies the complicated interactions between propri-etary, for-profit R&D and public “open science.”In this paper, we draw on experiences from thefield of genomics to examine some of the likelyconsequences of these interactions in neuroscience.

Facilitating the widespread sharing of data andtools for neuroscientific research will accelerate

the development of neuroinformatics. We propose approaches to overcome the cultural and legal bar-riers that have slowed these developments to date.We also draw on legal strategies employed by theFree Software community, in suggesting frame-works neuroinformatics might adopt to reinforcethe role of public-science databases, and proposea mechanism for identifying and allowing “openscience” uses for data whilst still permitting flex-ible licensing for secondary commercial research.

Disclaimer:This paper reflects the opinions and positions

of the authors and is not an official policy or opin-ion of any national government or organization.

149

*Author to whom all correspondence and reprint requests should be sent. E-mail: [email protected]

Neuroscience Data and Tool Sharing

A Legal and Policy Framework for Neuroinformatics

The OECD Working Group on Neuroinformatics: Peter Eckersley,a Gary F. Egan,*Shun-ichi Amari, Francesco Beltrame, Rob Bennett, Jan G. Bjaalie,Turgay Dalkara,Erik De Schutter, Carmen Gonzalez, Sten Grillner, Andreas Herz, K. Peter Hoffmann,Iiro P. Jaaskelainen, Stephen H. Koslow, Soo-Young Lee, Line Matthiessen, Perry L. Miller,Fernando Mira da Silva, Mirko Novak,Viji Ravindranath, Raphael Ritz,Ulla Ruotsalainen, Shankar Subramaniam, Arthur W.Toga, Shiro Usui, Jaap van Pelt,Paul Verschure, David Willshaw, Andrzej Wrobel, and Tang Yiyuanb

a Department of Computer Science & Software Engineering, and Intellectual Property ResearchInstitute of Australia, The University of Melbourne. E-mail: [email protected]. Eckersley’s work onthis project is supported by IPRIA funding from IP Australia.

b For the institutional affiliations of the other co-authors and members of the OECD working group onneuroinformatics, please see the end of this paper.

02/Egan149-166F 5/1/03 12:11 PM Page 149

150 _____________________________________________________________________OECD Working Group

Neuroinformatics _______________________________________________________________ Volume 1, 2003

IntroductionThe development of neuroinformatics

promises to extend several important trendsin scientific research into the practice of neu-roscience. The most visible is the integrationof data into large, international databases (andthe accompanying creation of a culture of datasharing1); another is the growth of research pro-jects beyond the scale of a few laboratories. Amore fundamental change is the virtualizationof the process of inquiry, enabling questions tobe answered with reference to secondary datarather than first-order empirical investigation.

Aside from the direct implications that thesechanges may have for the nature and scope ofscientific research, they also create importantquestions about policy and the organization ofinternational collaboration.

Perhaps the clearest precedent for the trans-formative effect neuroinformatics might haveon neuroscience can be found in the field ofgenomics. There, the Human Genome Project(HGP) and other international online collabo-rations have already come to fruition, and maycarry important lessons for neuroinformaticspolicy.

One of the most visible features of the HGPwas the entanglement between “open science”and the proprietization of information. Thewell-intentioned but simple strategy of plac-ing all HGP data in the public domain hasresulted in processes of proprietization—suchas the patenting of genes2 and the creation of

privately-held databases—from which theoriginal researchers receive little or no benefitand over which they have no control.

Certain “intellectual property” (IP) laws—or, more specifically, copyright, patent, anddatabase control laws—have played a grow-ing role in science over the last two decades(Dam, 1998). But the nature of electronic infor-mation access and exchange, in particular forlarge-scale, collaborative online research (suchas neuroinformatics) makes questions of infor-mation privatization more pressing and poten-tially problematic.

Small research projects can afford to set rulesthrough flexible negotiations which allow par-ticipants to obtain patents, for example, or topass technology into the public domain.

Large international collaborations, on theother hand, need to have clear policy groundrules, for a number of reasons. Firstly, negoti-ation between all parties is highly impractical,and in a complex research area, a thicket of pro-prietary patents (for example) can create “anti-commons” effects which hamper the progressof the field (Heller & Eisenberg, 1998).Furthermore, taxpayers and benevolent con-tributors require guarantees that the benefitsof research will flow as widely as possible,3 andfunding agencies require clear policies foraddressing these issues.

Whilst the short-term impact of patents andproprietary databases on open neuroscienceresearch may be relatively small, in the longterm, effects may be significantly larger.4

1 Although universal data sharing has been adopted in other fields such as physics and genomics, the issue remains controversial in neuroscience; see, for example (Aldhous, 2000; Koslow, 2000; Nature,2000; Koslow 2002).

2 Either as nucleotide sequences themselves, or indirectly through the proteins they code for.

3 For an extended discussion of the complicated interactions between the public interest and public pol-icy on the proprietization of research, see (Eisenberg & Rai, 2001).

4 An initial indication of the role of intellectual property privileges in the commercialization of neuroin-formatics is the Brain Resource Company (http://www.brainresource.com), whose promotionalmaterial states that they have obtained software and a “large quality controlled database of normativesubjects and with a range of clinical disorders,” and that they provide fee-for-service analysis reportsto clients including researchers, clinicians, and pharmaceutical companies, as well as for medico-legalpurposes.

02/Egan149-166F 5/1/03 12:11 PM Page 150

Volume 1, 2003 _______________________________________________________________ Neuroinformatics

Neuroscience Data and Tool Sharing __________________________________________________________151

Therefore, anticipating issues arising fromthe relationship between public and privatecontributions to neuroinformatic resources,and constructing appropriate policy frame-works from the outset is essential.

In this paper, we consider what regulations,guidelines, and organizations might comprisesuitable “policy frameworks” for internation-al neuroinformatics projects, and how theymight serve to balance the various competinginterests in neuroinformatics research. So, ratherthan being technical in a computer sciencesense, our proposals include a degree of legaltechnicality—suggesting ways in which the lawmight be used to provide infrastructure to fur-ther the progress of research in neuroscience.

The Current State of Neuroinformatics Policy Development

The Neuroinformatics Working Party of theOECD Megascience Forum (OECD, 1999)5

reported that, “the scientific goals of neuroin-formatics are to accelerate the progress of neu-roscience and informatics by: making betterand more efficient use of neuroscience datausing informatics-based, including computa-tional, approaches; generating and evaluatingnew hypotheses and computational theoriesabout brain function to drive further experi-ments; and developing and applying new toolsand methods for acquiring, visualizing, andanalyzing data important for understandinghow the brain functions.”

They proposed working towards these goalsby the implementation of a number of criticalsteps including: “Establish the coordination,

standardization and interoperability require-ments needed for successful application, inte-gration, stabilization and quality assessmentof the distributed and local neuroinformaticsfacilities; and enhance collaborative opportu-nities in neuroinformatics, both nationally andinternationally.”

The successful development of neuroinfor-matics will be dependent on achieving theseimplementation steps. Establishment of appro-priate software development and licensingmodels for sharing data and tools, as well asclarification of copyright issues pertaining tosharing publicly funded research results viainter-networked databases, is urgentlyrequired.

In the initial programs started in the UnitedStates under the aegis of the Human BrainProject (HBP), grantees were advised to pro-tect their newly developed informatics capa-bilities, databases, and analytical tools throughthe copyright mechanism, and then, where pos-sible, to openly share these resources. However,Grantees are also encouraged under this fund-ing, under the provisions of the Bayh-Dole Act,to patent and sell the products that are devel-oped from federally funded research.

In their recently published report on neu-roimaging databases, the international societyrepresenting the functional neuroimagingcommunity, the Organization for Human BrainMapping (OHBM) identified the followingquestions (OHBM, 2001):

1. What rules should govern the use of dataderived from a public database and who hasthe right to publish findings based on thesedata and within what time frames? Howshould credit be assigned?

5 The Organization for Economic Co-operation and Development (OECD) assists in the development ofmarket economies in its 30 member countries from the industrialized and emerging nations. TheOECD established the Global Science Forum (GSF) to foster cooperation in global large science pro-grams and issues, and recognizing the need for cooperative efforts in neuroinformatics, established aWorking Group on Neuroinformatics (WG-NI). The WG-NI included scientists and policy officialsfrom 24 of the OECD’s member and observer governments, and the recommendations of the groupwere recently published by the OECD at http://www.oecd.org/pdf/M00033000/M00033112.pdf.

02/Egan149-166F 5/1/03 12:11 PM Page 151



2. What rules should protect the confidential-ity of experimental participants, and howcan these be kept in alignment with localinstitutional review board regulations andinformed consent procedures?

3. What mechanisms should be implementedto prevent violations of these rules, whatrepercussions should ensue for infractions,and how can these enforced?

This discussion paper addresses these ques-tions and related questions concerning soft-ware tools used for neuroinformatics research.We consider the particularities of neurosciencecollaboration and draw on observations fromthe fields of genomics and free software devel-opment. We consider a number of licensingand software development models, and sug-gest possible legal and policy frameworks forthe neuroinformatics community.

Models for NeuroinformaticsSoftware Tool SharingAs identified by this group (Amari et al.,

2002), software tool sharing is one of the mostimportant aspects of neuroinformatic collabo-ration, since software plays such a large andcomplex role in neuroinformatics research.Unless researchers share the same softwaretools, it becomes more difficult to make pre-cise comparisons between results. This is notto say that all analyses must follow identicalprocedures, or that innovative new algorithmsand techniques should not be encouraged;rather, development should occur with a max-imal amount of transparency, extensibility, scal-ability, accuracy, and reproducibility. Softwaretool sharing models must navigate the legali-ties of copyright and liability in order to servethese goals.

The most well-known model for softwaredevelopment is the closed-source, proprietary

approach. Characteristically, a firm employsprogrammers and funds the development ofsoftware; copyright in their work is appropri-ated by contract, or work-for-hire clauses incopyright legislation. The company thenattempts to recoup its costs and make a profitby licensing their creations to customers.Although this model is very widespread andincludes examples such as Microsoft Windowsand Matlab, it has also been widely criticizedamongst software developers.6 Some of thebiggest weaknesses of this model involveinflexibility, centrally determined functional-ity, and sophisticated users’ loss of ability toexamine and extend source code, and to fixdefects in the tools they use. These problemsare particularly relevant to applications in sci-entific research.

Other weaknesses in the proprietary modelinvolve poor allocative and distributive eco-nomic efficiency (Shavell & van Ypersele, 2001),and a strong “network externality” effect7

which frequently leads to monopolies. In eco-nomic terms, software is a naturally non-rival-rous, non-transparent product; it is excludablebut only with significant loss of efficiency(DeLong & Froomkin, 2000). As a consequence,the software market is close to a classic exam-ple of “market failure.” There are a number ofother proprietary software development mod-els, such as shareware or application serviceprovider (ASP) approaches, but these also suf-fer from many of the weaknesses of the tradi-tional “sales” model.

Proprietary software development modelsare also poorly suited to collaborative tool shar-ing, since research groups with lower levels offunding may be unable to afford all softwarethat other researchers utilize. There is a strongeconomic incentive to specialize in the use ofa few packages, thus minimizing the number

6 For frequently cited examples, see Stallman, 1992 or Raymond, 1999.7 Where the commercial value of a product or tool increases with the number of other users.

02/Egan149-166F 5/1/03 12:11 PM Page 152



of required high-cost purchases. Thus, whennew techniques are developed, they receive alower degree of testing, scrutiny, and duplica-tion. Adifferent, but highly relevant, software devel-opment model is that of “free” or “open source”software. Free software is typically createdthrough collaborations over the Internet, withan ethos that is in many ways similar to thatof science, emphasizing the Mertonian normsof universalism, disinterestedness, commu-nalism,8 and skepticism (Merton, 1973). Thisculture plays an important (and growing) rolein the information technology (IT) industry.9

Examples of free software include GNU- andLinux-based operating systems, the Apacheweb server, and the GNOME and KDE desk-top environments. A very large amount of sci-entific software, such as Statistical ParametricMapping (SPM), Octave, or scientific Pythonextensions, have been developed using opencollaborative models. Examples of relevantindices of scientific software are available fromScientific Applications on Linux,10 bioinformat-ics.org,11 and Sourceforge.12

The main disadvantages of free softwaredevelopment models are related to fundingand incentives. Various theories are availableto explain the widespread creation of free soft-ware, relating to the prestige of authoring a

widely-used piece of software, and the oppor-tunities for cross-promotion (such as for con-sulting services).13 In general however, freesoftware development is an economic “publicgood,” and, as such, is subject to the “free-riderproblem,” where the number of people whouse the good is much larger than the numberwho contribute (financially or materially) toits production.

In a scientific research environment wherea substantial fraction of funding comes fromgovernment grants, this problem is circum-vented. The status quo is that many groupsdevelop free software tools as an aside to theirmain research focus; this practice benefits theentire international research community, andshould be encouraged.14 The limitation of themodel is that researchers may have insufficientincentive to document, offer support, and inte-grate such tools.

The copyright and liability issues involvedin this form of software development havealready been thoroughly explored, and thereare a range of free software licenses availablewhich give contributions clear legal status. So-called “copyleft”15 licenses may be the mostdesirable, in terms of guaranteeing that pub-lic funding for scientific software results inpublic benefit. There is also the additionaloption of dual-licensing software, which has

8 Which Merton, perhaps confusingly, referred to as “communism.”

9 See, for example: Shankland, S., “Linux sales surge past competitors,” CNET News, Feb 2000.(http://news.cnet.com/news/0-1003-200-1546430.html), or the Netcraft Web Server Survey(http://www.netcraft.com/survey/).

10 http://sal.rising.com.au11 http://bioinformatics.org12 http://sourceforge.net/softwaremap/trove_list.php?form_cat=9713

On the question of an economic explanation of free software production, see (Lerner & Tirole, 2000;Kuan, 2000; Kelty, 2001).

14Indeed, while DiBona et al., (2000) suggest that there are many similarities between free softwaredevelopment and science, Kelty (2001) claims that the free software movement is really a part of thescientific enterprise.

15 The term “copyleft” originated as a play on words from “copyright.” It refers to the use of powersgranted by copyright law, in order to provide benefits for the public, rather than copyright holders.These powers are chiefly used to constrain the private appropriation of derivatives of the copyleftwork (see http://www.fsf.org/copyleft for further details).

02/Egan149-166F 5/1/03 12:11 PM Page 153



potentially lucrative non-academic applica-tions. Dual (or “two tiered”) licensing, whichhas been used for successful software such asthe QT GUI framework16 and the MySQLdatabase,17 allows otherwise copylefted toolsto be sold to firms that wish to develop pro-prietary systems with them. The success of thefree software movement in creating widelyused and successful software tools by collab-oration over the Internet sets a useful prece-dent for the field of neuroinformatics.

At the level of individual laboratories andresearch projects, we urge scientists to releaseany software they are developing under freesoftware licenses (such as the GNU GeneralPublic License), and, where possible, to choosefree software tools for their work.

At a broader policy level, we suggest that agood strategy for encouraging software toolsharing would be to allocate resources for theapplication of larger-scale software engineer-ing practices and resources to pre-existing freescientific tools. Such a policy would result ina more standardized, more organized array oftools for international electronic collaboration.There is a role for both national science insti-tutions and international neuroinformaticsbodies to encourage and organize funding forthese activities.

Models for NeuroinformaticsDatabase SharingAs observed by Amari et al. (2002), the com-

plexity, heterogeneity, and contextual natureof neuroscience data means that the technicaltask of creating neuroinformatics databases isdaunting. Nonetheless, there are also signifi-cant scientific cultural and political obstaclesto primary data sharing that slow the con-struction of such databases. This is a conun-drum, because the universal use of databases

would provide more data and stronger incen-tives to accelerate neuroinformatics databaseresearch. Once shared database technology andcontent has matured, it is realistic to expect thatpolicies such as those adopted by the genomicscommunity (DOE-NIH, 1993) would beinevitable.

The experience of bioinformatics has alsoshown that the IP implications of researchdatabases are quite profound. The promise ofgenetic engineering has seen public researchdata used as the starting point for numerousprivate commercial endeavors. Corporationshave built proprietary datasets using publicdomain research databases to improve theirinformation (Eisenberg, 2000); patents havebeen claimed on uses for genetic informationthat publicly funded researchers have placedin open databases. This environment has cre-ated significant consternation and debateamongst researchers, lawyers, and the widerpublic; it is uncertain whether a successful bal-ance will be found to match incentives for R&Dagainst the public interest in unfetteredresearch and open markets.

Indeed, the potential for bioinformatic tech-niques to seed a “thicket” of proprietary rightsthat inhibit an entire field of research has beenrecognized not only by academics and policymakers, but also by large pharmaceutical firms.The work of creating a public domain databaseof Single Nucleotide Polymorphisms (SNPs)is being undertaken by the SNPs Consortium,which is funded by “big pharma” and whichemploys complicated and non-obvious strate-gies to ensure that as much information as pos-sible remains unencumbered by patent laws(Marshall, 1997; Eisenberg, 2000; Sunder Rajan,2002). It is possible that similar strategies todelineate the public domain in neuroinfor-matics will prove valuable.

16 http://www.trolltech.com

17 http://www.mysql.com

02/Egan149-166F 5/1/03 12:11 PM Page 154



As with software, one of the most importantpolicy instruments available for guiding theoperation of collaborative databases are sug-gested or prescribed licensing models foraccess to these databases. There are a numberof means by which such models can determinewho is able to access what data, and how theycan use that information as the basis for newworks.

Database Licensing Models

Databases, in general, attract varying formsof protection under national copyright laws;the status of these laws is complicated, but insome jurisdictions, copyright may subsist inthe aggregation of content in a database. In theUnited States, courts have ruled that there canbe no copyright in a database of “mere facts”(Feist, 1991), while in Europe, regardless of thestatus of databases under national copyrightlaws, the EU has enacted specific sui generis(Latin: “in its own class”) database rights. Thereis significant debate as to whether extensive IPprivileges in databases are desirable.18 Fordetailed discussions of the interaction betweenevolving database policy and open science, seeReichman & Uhlir, 2001 and David, 2001.

The situation for neuroinformatics databas-es is, however, quite different from that ofdatabases in general. Unlike simple collections,such as the telephone directory in Feist vs RuralTelephone (Feist, 1991), which could be said tocomprise “mere facts,” neuroscientific dataincludes a great deal of material in which copy-right would subsist directly, such as magneticresonance images. As a result, copyright licens-es applied to much of the content in a neu-roinformatic database would be practicallyenforceable in a wide range of jurisdictions.

A second instrument of collaborativeresearch policy is the requirement of a con-

tractual agreement for use of a database. Celerauses agreements such as these to protect itscommercial interests,19 but researchers mightchoose to employ them in the service of an“open science” model; imposing certainrequirements for non-privatization of knowl-edge might serve both the public interest andthe scientific community.

Finally, there is the possibility of using elec-tronic access controls to protect databases. Inpractical terms, such controls might only beused to support contractual access (requiringagreement before granting accounts) and toprotect database quality (requiring authenti-cation before allowing data to be deposited).Stronger access controls, such as Digital RightsManagement systems would be both extreme-ly difficult to implement and inappropriate forconsideration in a scientific setting.

There are a number of broader objectiveswhich all of the licensing models shouldattempt to achieve; allowing access to theresearch community, private corporations andthe public; guaranteeing scientists credit fortheir contributions; preventing the uncom-pensated privatization of public knowledge;and maintaining incentives for commercialR&D. We now consider a number of databaselicensing models which might be adopted, andevaluate them by comparison with these goals.

Public Domain Databases

This is the traditional model adopted by“open science,” which provides free publicaccess to information and is suitable for sci-entific collaboration. However, it may be dis-advantageous for researchers, because it doesnot guarantee them credit for their data underall circumstances, and does not provide themwith any reward if extensions of their work arecommercialized.

18 See, for example, the report of the US National Research Council (NRC, 1999), (Steele, 1996) or(Benkler, 2000).

19 See Celera Genomics, Terms & Conditions for use, http://www.celera.com/company/terms_conditions.cfm.

02/Egan149-166F 5/1/03 12:11 PM Page 155



Databases Covered by Copyright Law (to the extent permissible by nationallaws, but with open access grantedthrough a copyleft license)This model retains all the benefits of public

domain databases. It has the additional advan-tage that scientists will be in a position to nego-tiate for royalties in a situation where a privatefirm wishes to make proprietary extensions toa database (since they will need a special licenseagreement). However, it provides no “guar-antee of credit” to the submitters of data, pro-vides only weak protection against “patentthickets,” and grants scientists no reward forpatents on extensions to their work.

Contractual Access to CopyrightedDatabasesIn this model, the database is covered by

copyright with a copyleft license, but in addi-tion, access to the database is provided only toaccount holders who have agreed to a contractwith the database administrators. This contractwould include a number of relevant provisions:a credit allocation requirement (see “Databases,Authorship, and Credit”). The user is prohib-ited from creating proprietary derivations,such as new databases, software or patents,using the database. If a user wishes to do thesethings, they may negotiate a special contractthat allows them to do so. The user is prohib-ited from passing information from thedatabase to other parties for the purpose of cir-cumventing the other terms of the contract. Notethat redistribution of the data is otherwise per-mitted by the copyleft license.

A contractual arrangement of this sortwould, in practice, provide a very good guar-antee of credit for researchers. It does not pro-vide a watertight guarantee of reward in thecase of patents arising from secondary research,since once a third party obtains the data, they

are not actually bound by the terms of the agree-ment (this could only be achieved by a far morestringent contract which would significantlycomplicate some legitimate scientific uses).Nonetheless, it does provide a significantdegree of protection, and commercial userswould in general be unwilling to risk plannedcircumvention.

Administering Databases Using aCombined Contractual-Copyleft Model with Dual-LicensingThe “contractual access to copyrighted

databases” model would add an additional(small) degree of complexity in administeringdatabase access, because under most jurisdic-tions, contracts are only enforceable wherethere is clearly demonstrated agreement,understanding, and negotiability (Burk, 2000).Ascientific database might be able to meet thesecriteria by means of the following:

• Allowing a prospective user to request anaccount subject to the agreement;

• Asking a few simple questions about thenature of the contract;

• Offering an alternative channel for negotia-tions if they would prefer different terms(such as a fee-based agreement allowing useof the database in developing proprietarypatents).

Such a system would be analogous to the“dual-licensing” model used by some free soft-ware developers and provides a very elegantmechanism for distinguishing between “pub-lic good” and commercial uses,20 but there arestill a number of unresolved issues to consid-er. Would proprietary licenses be available bynegotiation with the organization administer-ing the database, or only by direct discussionwith the original researchers? In the formercase, royalties given to the administrative body

20A recent paper by Reichman and Uhlir comes to many of the same conclusions about database licensingas we do; their work attempts to resolve the tension between open science and commercializability byanalogizing the GNU Lesser General Public License (http://www.gnu.org/copyleft/lesser.html), aweak form of copyleft (see particularly (Reichman & Uhlir, 2001) p. 320). We believe, however, that thedual licensing model provides stronger protection for the public interest, clearer delimitation betweenuse cases, and greater certainty that researchers will be remunerated when their work is commercial-

02/Egan149-166F 5/1/03 12:11 PM Page 156


would need to be redistributed back to theresearchers, perhaps through some grantingstructure. In the latter case, unless client firmscould find a practical way of negotiating withthe researchers as a group, proprietary exten-sions of a large fraction of the database mightbe difficult to arrange. Whether such arrange-ments might be desirable is a complex policyquestion in and of itself.

Accountability for Collective LicensingOrganizationsIf the administrators of a database or their

funding agencies are given responsibility fornegotiating the terms for proprietary licenses(we will henceforth refer to such organizationsas Collective Licensing Organizations, orCLOs), there are a number of potential “moral

hazards”21 to be addressed, and a need existsfor effective guarantees of accountability. Inparticular, scientists need confidence that theirinterests will be securely represented in anysuch negotiation.

Guarantees of CLO accountability to scien-tists are unlikely to be robust if all rights to dataare automatically assigned to the CLO. In orderto make accountability systematic, we proposea mechanism in which there is collective licens-ing, but individual researchers or laboratoriescan withdraw if they believe their interests arenot being represented.22

This “collective licensing with opt-out”scheme allows efficiencies of scale in negotia-tion, but preserves the sovereignty of each con-tributor (see Fig. 1). We believe this addresses

Fig. 1. Structure of the proposed copyleft/dual-licensing arrangement for neuroinformatics databases. Colorfigure available online.

21 Moral hazards are said to occur when actors have net incentives to act contrary to their appropriaterole in an agreement (or organization).

22 It is realistic to expect that different contributors will have quite different expectations about thelicenses issued on their behalves. Some might seek a maximum rate or return from commercialization,while others might wish to guarantee flows of knowledge back to open science, or apply various ethi-cal criteria to secondary research. For a CLO, the logical approach to this problem would be to catego-rize different positions, and record these as metadata annotations in the relevant databases.

02/Egan149-166F 5/1/03 12:11 PM Page 157



many of the potential “institutional design”pitfalls associated with creating CLOs.

Databases,Authorship, and CreditOne of the concerns widely regarded as an

obstacle for collaborative data sharing is thatof credit for one’s data. Many researchers feelthat if they publish their primary experimen-tal results along with their first paper, betterresourced groups might pre-empt further pub-lications. Advocates of data sharing have sug-gested that the additional expertise of the orig-inal researcher with regard to this data wouldbe sufficient to prevent this from being prob-lematic (Koslow, 2000; Nature, 2000).Furthermore, these analyses may underesti-mate the incentives that primary data publi-cation would create for forming collaborations,to the direct benefit of the original author. TheOHBM has identified credit allocation as a keyquestion to be addressed for sharing neu-roimaging data (OHBM, 2001). There are anumber of different approaches that could betaken, including exclusion, citation, and co-authorship.

ExclusionThe original author has a fixed or negotiat-

ed “window of privilege” during which otherresearchers may not publish using that data.This has the serious disadvantage of slowingand complicating the research process, andwould act as a disincentive for distributed col-laboration.

Requirement of CitationA citation must be made to the original

researcher’s paper. One problem is that somedata may not have an associated publicationto cite; this could be solved through the cita-tion of special technical reports describingexperimental techniques.23

Co-Authorship

Secondary data users could be required tooffer co-authorship to the original researcher.Some journals have authorship requirementsthat might make this difficult, but in manycases, this approach could foster stronger col-laboration as well as allocating credit andwould enhance the re-analysis of the data ifthe original data producer was to be involvedin the new interpretation.

Such policies could be made a condition ofaccess through database licenses. On balance,it would appear that the third option, perhapsfalling back to the second in situations wherejournal policies are incompatible, would bestserve the interests of the neuroinformaticsresearch community. Other options to be con-sidered could be a separate citation byline onthe title page listing the data source and theoriginal authors.

Enforcing License Conditions

The OHBM (OHBM, 2001) identifiesenforcement of access rules as an importantissue to be addressed. Organizations such asthe Free Software Foundation already exist inpart to provide legal support for softwaredevelopers who might need to prosecute vio-lations of free software copyright licenses.Collaborative neuroinformatics software mayobtain some degree of protection from suchorganizations.

If more specialized contract/copyleft accessmodels were deployed for neuroinformaticdatabases, the organization administering thedatabase would probably have to take on someresponsibility for preventing misappropria-tion. The experience of the free software worldhas been, however, that when the community“cries foul” over license violations, the situa-tion is often rectified before legal action is con-sidered.24

23 Preferably, a scientific paper should have been published describing all of the data which is insertedinto neuroinformatics databases; in cases where this is not possible, at least a technical report describ-ing experimental techniques should be published.

02/Egan149-166F 5/1/03 12:11 PM Page 158



Privacy Issues in NeuroinformaticsRestrictions on the Transfer

and Processing of Personal and Sensitive DataResearch projects have obligations (both

legal and ethical) to guarantee the privacy oftheir subjects. This is especially true in fieldswhere sensitive information about a large num-ber of people is collected in well-indexeddatabases.

In some countries, privacy laws are relativelyweak; the United States, for example, requiresonly that organizations collecting personalinformation provide citizens with the abilityto “opt-out” of having those organizationsshare their data with third parties. At a globallevel, however, privacy regulations are expand-ing rapidly (White & Case, LLP, 2002), and itis probably wise to base policy analysis forinternational collaboration on the strongestapplicable laws.

The European Union, for example, placesstrict constraints upon the sharing of “sensi-tive data,” which includes data about racial orethnic origin, and hence would probablyinclude much of the information in neurosci-entific databases.25 In general, such informa-tion can only be processed by neuroinfor-

maticians for non-clinical purposes if the sub-ject has given their explicit consent to this pro-cessing.26 The EU also restricts the export ofdata to states which do not have strong priva-cy laws, although the use of appropriate con-tracts will allow exports to parties which agreeto abide by EU-equivalent regulations.27

Individual research groups, or nationalresearch administrations, could facilitate datasharing which is compatible with privacy lawsworldwide, by obtaining informed consentfrom experimental subjects, for the use of(anonymized) sensitive data about them, forresearch purposes. The process of drafting andemploying these consent agreements at thelaboratory or even the national level may beslow, however, because of the need to ensurecompatibility between privacy standards andthus enable the exchange of data.

Our suggestion is that steps should be takento draft and employ consent agreements at aninternational level, which are compatible withstrong privacy laws such as those found in theEU. Such a program dovetails well with ourrecommendations about database licensing,for a number of reasons. Firstly, there is alreadya contractual framework in place for access tothe databases, within which privacy rules canbe included. Secondly, the dual-licensing

24 A small number of copyleft license violations have been identified by people examining proprietaryexecutable programs and identifying similarities between these and GPLed software. For some exam-ples, see http://slashdot.org/search.pl?query=GPL+violation&sort=2 . It seems unlikely that signifi-cant violations could go unnoticed for long periods of time, especially within a scientific researchcommunity. Moglen (2001) describes the process and success of GPL enforcement in more detail.

25 There are slightly weaker requirements for data which is personal but does not fall into the “sensi-tive” category; however, considering the fact that neuroscientific data will often be very revealing,and that some nations may apply strict regulation to both classes of data, our analysis follows themore restrictive case.

26 See Article 8 of EU privacy directive (European Parliament, 1995). The exception given in Article 8, §3,which covers clinical uses of data, is probably not broad enough to allow neuroinformatic research.Also note that the directive does allow European states to enact national laws which both grant addi-tional permissions for various uses of data, and laws which limit the extent to which individuals cangive consent for certain uses.

27 See the European Commission Decision on standard contractual clauses for the transfer of personaldata to third countries,http://europa.eu.int/comm/internal_market/en/dataprot/news/clauses2.htm

02/Egan149-166F 5/1/03 12:11 PM Page 159



scheme fits well with the requirement ofinformed consent by subjects; if licensing isclearly separated between non-proprietary,research purposes, and separately licensedcommercial uses, the subjects can grant con-sent to either or both of these classes of usage.

Provided the meanings of the differentlicensing cases are succinctly explained, andemphasis is placed on the fact that data will beanonymized and legally protected, it is likelythat most people will grant permission for bothforms of neuroinformatics. Furthermore, thosesubjects who are unwilling to allow privateR&D with their data will still be able to con-tribute to open scientific research.

Anonymization of Neuroscientific DataAny open, large scale, international databas-

ing of neuroscientific data must necessarily bebased on the principle that, although intimateand sensitive information is available aboutnumerous individuals, this information isinsufficient to be linked to their identities.

Thus, although the database might includeanatomical, demographical, and pathologicalinformation, it should never contain names orlinks to other databases which contain identi-fying information.

It should be noted that some neurosciencedata types, such as MRI images, contain inher-ently identifying information about facial andcranial structure.28 Attempts could be made toperform cropping or transformation, so as toremove the unique link to the subject’s identi-ty. An extended approach might be to holdentire raw datasets in escrow, but make onlydefined subsets directly available through thedatabase. If it transpired that the available sub-sets were insufficient for required analyses,

then alternative access to the full dataset couldbe negotiated.

Beyond the point of removing bone struc-ture and registering neuroanatomy, such astrategy, however, becomes more problematicbecause it is unlikely that transformationscould absolutely conceal identities without vir-tually destroying all structure of scientific inter-est. Fortunately, it may not be necessary,because although transformed brain tissueimages are inherently identifying, they onlypossess this property when the party seekingto identify a subject already has knowledge ofthe anatomy and pathology of particular indi-viduals. This fact appears to constrain the pri-vacy problems inherent in neuroinformaticresearch.

So then, the key challenges for guarantee-ing anonymization in neuroinformaticsdatabases are ensuring that no combinationsof data are stored which together become iden-tifying, and that any potential attempt at iden-tifying a subject would have to go to inordi-nate lengths—such as obtaining brain scansfrom other sources—in order to recognize aparticular individual. Once these steps aretaken, subjects should be informed of theseprinciples when consenting to the use of theirdata.

A recent example of a neuroimagingdatabase partially addressing these issues isthe fMRI Data Center which has been estab-lished, “to help speed the progress and theunderstanding of cognitive processes and theneural substrates that underlie them by: pro-viding a publicly accessible repository of peer-reviewed fMRI studies; providing all data nec-essary to interpret, analyze, and replicate thesefMRI studies.”29 This is a public domain

28 Although the authors are not aware of any tools for identifying individuals based on MRI data, thesignificant body of biometrics literature and software for face recognition from video, would suggestthat such identification would be achievable. In particular, it is likely that cranial structure could belinked to photographs, and that neuroanatomy could serve as a unique “fingerprint” for identifyingindividuals.

02/Egan149-166F 5/1/03 12:11 PM Page 160



database, which requires users who wish topublish data to credit the authors of the origi-nal study and acknowledge the Center and theaccession number of the dataset. The Centerrequires contributors to remove subject iden-tifying tags, and the Center also “anonymizes”high resolution structural MR data by remov-ing non-brain tissues. However, the Centerdoes not require users to explicitly agree to thecontractual terms of access.

SummaryAlthough the complexity and heterogeneity

of neuroinformatics make the establishment ofwidespread collaboration a daunting task, thepotential benefits to the field are huge. Sharingdata and tools are perhaps the most importantaspects of online collaboration; without thissharing, significant costs and effort recur inevery research group.

The scientific R&D community has, in recentyears, been pulled in two directions by oppos-ing forces. On one hand, the traditional ethosof “open science” has suggested that informa-tion and tools should be shared, and that coop-eration should be encouraged as widely as pos-sible. On the other hand, the pressures of fund-ing requirements, governmental policy, andmarket opportunities have increasinglyrequired scientists to think in competitiveterms, seeking to commercialize their work.The presence of commercial incentive struc-tures at times appears to be irreconcilable with“open science.” In this environment, the mostvaluable knowledge is kept secret or made pro-prietary.30

With sufficient foresight and planning, how-ever, it is not impossible for the open scientif-ic community to obtain the collaborative ben-efits of “open science” whilst simultaneously

appropriating some of the economic value ofits creations. This could be achieved by simul-taneously adopting the copyleft principle inorder to maximize the free flow of collabora-tive information and allow for negotiated alter-native commercialization licenses for the pri-vate sector. Adopting policy positions thatencourage the open sharing of data and tools,and the use of legal frameworks which sup-port “open science” collaborations, are themost effective strategies available for nurtur-ing the field of neuroinformatics.

AcknowledgmentsThanks to David Brennan, Jeff Bizzaro,

Richard Stallman, and David Lindsay for theirhelpful information and feedback on some ofthe ideas discussed in this paper.

Appendix 1: Glossary of TermsApplication Service Provider (ASP): A firm

which services the outsourced ITneeds of another organization. Thiswould typically include runningsoftware on the ASP’s computers toprovide services for their clients.

Copyleft: A free software license whichrequires modified versions of thesoftware to have the same license.Thus, copyleft software in somesense belongs to the public, and pri-vate entities do not have the right tocreate privatized derivatives. Themost commonly used form of copy-left is the GNU General PublicLicense, or GPL.

Dual Licensing: The practice of releasingsoftware under several licensessimultaneously. Copyright holdersmay choose to use any combination

29 http://www.fmridc.org/about30 For further discussion of this process, and its relationship to IP privileges, see Rai, 1999 and David,

2001.

02/Egan149-166F 5/1/03 12:11 PM Page 161



of free (copyleft or otherwise) andproprietary licenses.

Executable Code: The form of software that isused on a computer when a programruns. It is created from the source codeby a compiler program. Proprietarysoftware is usually distributed in exe-cutable form only.

Free Software: Allows users a significant num-ber of freedoms including the right toaccess the software’s source code, andthe right to reproduce, modify, anddistribute the software. Free softwareis typically developed in highly col-laborative communities on theInternet. Free software comes in twovarieties, copyleft software and non-copyleft free software. Unlike publicdomain software, free software is oftenprotected by copyright laws.

Freeware: An ambiguous term for software that isavailable at no cost. Often, freeware is pro-prietary software available only in exe-cutable form. Thus, although freeware maybe non-commercial, the author still retainscontrol over the distribution and functionof the program.

GNU General Public License (GPL): A strongform of copyleft; it applies to the entire-ty of larger works which are derivedfrom the original. The GPL thus pre-vents libraries from being linked toproprietary applications.

Non-copyleft Free Software: Free software towhich proprietary extensions can bemade.

Open Source Software: see free software.Proprietary Software: Software that is owned

(and controlled) by a single, usuallycommercial, entity.

Public Domain Software: Software which isnot covered by copyright law; thus,anybody is able to do anything with it.

Shareware: Aform of freeware where a restrict-ed version of the software is madeavailable at no cost; a fully-featuredversion is available from the authorfor a fee.

Source Code: The form in which programmerscreate software. It is written in a par-ticular programming language, suchas C++, Pascal or Java. It is generallyfeasible for programmers to change aprogram if they have access to thesource code.

References

Aldhous, P. (2000) Prospect of data sharing givesbrain mappers a headache. Nature 406:445–446.

Amari, S., Beltrame, F., Bjaalie, J., et al. (OECDNeuroinformatics Working Group) (2002).Collaborative Neuroscience: Neuroinformatics forSharing Data and Tools. J. Integ. Neuro. 1:117–128.

Benkler, Y. (2000) Constitutional Bounds ofDatabase Protection: The Role of Judicial Reviewin the Creation and Definition of Private Rightsin Information. Berkeley Technology Law Journal15:535–595.

Burk, D. L. (2000) Intellectual Property Issues inElectronic Collaborations. In: ElectronicCollaboration in Science (Koslow, S. H. and Huerta,M. F., eds.) Lawrence Erlbaum Associates,Mahwah, NJ, pp. 15–44.

Dam, K. (1998) Intellectual Property and theAcademic Enterprise.John M. Olin Law &Economics Working Paper 68, University ofChicago Law School, Chicago, IL.

David, P. (2001) Will Building “Good Fences” ReallyMake “Good Neighbours” in Science? In: IPR(Intellectual Property Rights) Aspects of InternetCollaborations (Granstrand, O., Foray, D., &David, P., eds.) European Commission ResearchDirectorate General. Brussels, Belgium.

DeLong, J. B. and Froomkin, A. M. (2000)Speculative Microeconomics for Tomorrow’sEconomy. First Monday 5(2) http://firstmonday.dk/ University of Chicago, Chicago, IL.

DiBona, C., Ockman, S., and Stone, S., eds. (2000)Introduction, Open Sources: Voices from theOpen Source Revolution, O’Reilly & Associates,Sebastapol, CA.

02/Egan149-166F 5/1/03 12:11 PM Page 162



DOE-NIH Guidelines for Sharing Data andResources (1993) Human Genome News 4(5) 4.

Eisenberg, R.A. (2000) The Public Domain inGenomics. Proc. “A Free Information Ecology inthe Digital Environment” conference, New YorkUniversity School of Law Information Institute,New York, NY.

Eisenberg, R. A. and Rai, A. K. (2001) The Publicand the Private in Biopharmaceutical Research.Proc. Conference on the Public Domain, DukeUniversity, Durham, North Carolina, pp.157–175.

European Parliament and of the Council of theEuropean Union (1995). Directive 95/46/EC onthe protection of individuals with regard to theprocessing of personal data and on the free move-ment of such data. European Commission,Brussels.

Feist Publications vs. Rural Telephone ServiceCompany Inc., 499 U.S. 340, 1991.

Heller, M. A. and Eisenberg, R. S. (1998) Can PatentsDeter Innovation? The Anticommons inBiomedical Research. Science 280: 698–701.

Kelty, C. (2001) Free Software/Free Science. FirstMonday 6(12) http://firstmonday.dk/;University of Chicago, Chicago, IL.

Koslow, S. H. (2000) Should the neuroscience com-munity make a paradigm shift to sharing pri-mary data? Nature Neuroscience 3:863–865.

Koslow, S. H. (2002) Sharing primary data: a threator asset to discovery? Nature Reviews/Neuroscience3:311–313.

Kuan, J. (2000) Open Source Software As ConsumerIntegration into Production. Unpublished, avail-able from http://freesoftware.mit.edu/online_papers.php

Lerner, J. and Tirole, J. (2000) The simple economicsof open source. NBER working paper 7600.National Bureau of Economic Research,Cambridge, MA.

Marshall, E. (1997) Snipping away at genomepatenting. Science 277:1752–1753.

Merton, R. K. (1973) The Sociology of Science:Theoretical and Empirical Investigations,Chicago University Press, Chicago, IL.

Moglen, E. (2001) Free Software Matters: Enforcingthe GPL, parts I & II. Linux User, August-September 2001, Cheshire, UK, pp. 66–67.

Nature Neuroscience editorial (2000) NatureNeuroscience 3:845–846.

OECD (1999) Final report of the OECD Megascienceforum—Working Group on Biological Informatics,OECD Publications. http://www.oecd.org/pdf/M000014000/ M00014759.pdf, Paris, France.

Governing Council of the Organization for HumanBrain Mapping (OHBM) (2001) NeuroimagingDatabases. Science 292:5522:1673–1676.

Rai, A. K. (1999) Regulating Scientific Research:Intellectual Property Rights and the Norms of Science.Northwestern University Law Review 4(1): 77–152.

Raymond, E. S. (1999) On management and theMaginot Line. In: The Cathedral and the Bazaar.http://www.tuxedo.org/~esr/writings/cathe-dral-bazaar/cathedral-bazaar/x340.html,O’Reilly & Associates, Sebastapol, CA.

Reichman, J. H. and Uhlir, P. F. (2001) PromotingPublic Good Uses of Scientific Data: AContractually Reconstructed Commons forScience and Innovation. Proc. Conference on thePublic Domain, Duke University, Durham, NC,pp. 239–322.

Shavell, S. and van Ypersele T. (2001) Rewards ver-sus Intellectual Property Rights. J. Law. Econ.44(2):525–548.

Stallman, R. M. (1992) Why Software Should beFree. http://www.gnu.org/philosophy/shouldbefree.html. Free Software, Free Society:Selected Essays pf Richard M. Stallman (2002),Guy, J. ed. GNU Press, Boston, MA.

Steele, S. (1996) Electronic Frontier Foundationcommentary on the “Sui Generis Protection ofDatabases” treaty proposed for the DiplomaticConference on Certain Copyright andNeighboring Rights Questions in Geneva.Unpublished but available at http://www.eff.org/Intellectual_property/eff_wipo_19961122.comments.

Stewart, J. E., Mangalam, H., and Zhou, J. (2001)Open Source Software meets Gene Expression.Brief Bioinform 2:319–328.

Sunder Rajan, K. (2002) Banking (on) Biologicals:Commodification and Global Circulation ofHuman Genetic Material. Sarai Reader 2:277–289.

US National Research Council (NRC) (1999) Aques-tion of balance: Private Rights and the PublicInterest in Scientific and Technical Databases,National Academies Press, Washington, DC.

White & Case, LLP (2002) Global Privacy Law: ASurvey of 15 Major Jurisdictions. Global PrivacyForum. http://www.whitecase.com/report_global_privacy.pdf.

02/Egan149-166F 5/1/03 12:11 PM 3



OECD Neuroinformatics WorkingGroup Members’ Affiliations:

AustraliaGary F. EganHoward Florey InstituteUniversity of MelbourneParkville 3010,AustraliaEmail: [email protected]

BelgiumErik De SchutterBorn Bunge FoundationUniv Antwerp-UIAEmail: [email protected]

ChinaTang YiyuanInstitute of NeuroinformaticsDalian University of TechnologyEmail: [email protected]

Czech RepublicMirko NovakFaculty of TransportationCzech Technical UniversityInstitute of Computer ScienceAcademy of Sciences

of the Czech RepublicLaboratory of System ReliabilityEmail: [email protected]

Vaclav SebestaInstitute of Computer ScienceAcademy of Sciences of the Czech RepublicEmail: [email protected]

European CommissionLine MatthiessenScientific OfficerEuropean CommissionDG Research B-II-03 - NeurosciencesE-mail: line-gertrud.matthiessenguyader @cec.eu.int

FinlandIiro P. JaaskelainenMassachusetts General Hospital-NMR CenterHarvard Medical SchoolEmail: [email protected]

Dr Ulla RuotsalainenTampere University of TechnologyDMI/ Signal Processing LaboratoryEmail: [email protected]

GermanyAndreas V. M. HerzHumboldt-Universitaet BerlinInnovationskolleg Theoretische BiologieTheorie neuronaler SystemeEmail: [email protected]

Klaus-Peter HoffmannUniversitaet BochumAllgemeine Zoologie und NeurobiologieEmail: [email protected]

Raphael RitzHumboldt-Universitaet BerlinInnovationskolleg Theoretische BiologieTheorie neuronaler SystemeEmail: [email protected]

IndiaViji RavindranathOfficer on Special DutyNational Brain Research CentreICGEB CampusEmail: [email protected]

ItalyFrancesco BeltrameDelegate from the Italian Ministry

of Universityand of Scientific andTechnological ResearchEmail: [email protected]

JapanShun-ichi AmariVice DirectorRIKEN Brain Science InstituteLaboratory for Mathematical NeuroscienceEmail: [email protected]

Shiro UsuiBiol. & Physiol. Eng. Lab.Dept of Inf. & Comp. Sci.Toyohashi Univ. of TechnologyEmail: [email protected]

02/Egan149-166F 5/1/03 12:11 PM Page 164



KoreaSoo-Young LeeDepartment of BioSystemsDirector, Brain Science Research CenterKorea Institute of Science and TechnologyEmail: [email protected]

NetherlandsJaap van PeltNetherlands Institute for Brain ResearchMeibergdreef 33, 1105 AZ AmsterdamThe NetherlandsEmail: [email protected]

NorwayJan G. BjaalieNeural Systems & Graphics ComputingDepartment of AnatomyInstitute of Basic Medical SciencesUniversity of OsloEmail: [email protected]

PolandAndrzej WrobelProfessor of NeurosciencesNencki Institute of Experimental BiologyEmail: [email protected]

PortugalFernando Mira da SilvaINESC - Grupo de Redes NeuronaisEmail: [email protected]

SpainCarmen GonzalezDepartamento de FarmacologíaFacultad de MedicinaUniversidad Miguel HernándezEmail: [email protected]

SwedenSten GrillnerNobel Institute for NeurophysiologyDep. NeuroscienceKarolinska instituteEmail:[email protected]

SwitzerlandPaul VerschureInstitute of Neuroinformatics, ETH-UZEmail: [email protected]

TurkeyTurgay DalkaraProfessor of NeurologyInstitute of Neurological Sciences

and PsychiatryHacettepe University and Advisory

to the PresidentScientific & Technical Council of TurkeyEmail: [email protected]

United KingdomRob BennettBoard Programme ManagerMRC Neurosciences & Mental Health BoardMedical Research CouncilEmail: [email protected]

David WillshawInstitute for Adaptive and Neural ComputationDivision of InformaticsUniversity of EdinburghEmail: [email protected]

United StatesStephen H. Koslow (Chairman)Associate DirectorNational Institute of Mental Health, NIHDirector, Office on NeuroinformaticsEmail: [email protected]

Perry L. MillerDirector, Center for Medical InformaticsYale University School of MedicineEmail: [email protected]

Shankar SubramaniamProfessor of BioengineeringChemistry and Biochemistry, UC San DiegoEmail: [email protected]

Arthur W. TogaLaboratory of Neuro Imaging4238 Reed BldgDepartment of NeuorologyUCLA School of MedicineEmail: [email protected]

02/Egan149-166F 5/1/03 12:11 PM Page 165

02/Egan149-166F 5/1/03 12:11 PM Page 166

Neuroinformatics - Universitetet i oslo · Neuroinformatics _____ Volume 1, 2003 Introduction The...

Documents

Transcript of Neuroinformatics - Universitetet i oslo · Neuroinformatics _____ Volume 1, 2003 Introduction The...